Publication: Enhanced Filter-Wrapper Feature Selection Using Ensemble And Metaheuristic Approaches For High-Dimensional Multi-Class Text Classification
| dc.contributor.author | Peter, Ige Oluwaseun | |
| dc.date.accessioned | 2026-06-19T03:19:13Z | |
| dc.date.available | 2026-06-19T03:19:13Z | |
| dc.date.issued | 2025-05 | |
| dc.description.abstract | This study addresses the challenge of high-dimensional multiclass text classification, where traditional feature selection struggles with local optima and feature interaction problems. While filter methods are efficient, they suffer from bias and instability. Ensemble approaches help but have limitations, prompting the need for hybrid techniques. The study introduces a filter-wrapper approach, combining multi-univariate ensemble feature selection (munifes) with a hybrid artificial bee colony (abc) and genetic algorithm (ga) optimization. Munifes combines chi-square, anova, and information gain multi-univariate features using weighted concatenated voting to select discriminative features across 10 classifiers. The wrapper method integrates abc and ga with munifes to enhance feature selection using ann-adjusted mutation rates, tournament selection, and employed bee indexing for exploration-exploitation balance. Experiments on 20 newsgroup and 17 newsgroup datasets show over 50% feature reduction while achieving 96% and 100% classification accuracy, outperforming state-of-the-art methods. These results confirm the approach’s effectiveness in preserving population diversity and enhancing text classification performance. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/24420 | |
| dc.language.iso | en | |
| dc.subject | Image Processing Image Processing Equipment Industry Image Processing--Congresses Image Processing--Digital Techniques Image Processing--Digital Techniques--Congresses Image Processing--Digital Techniques--Handbooks | |
| dc.subject | Manuals | |
| dc.subject | Etc. | |
| dc.title | Enhanced Filter-Wrapper Feature Selection Using Ensemble And Metaheuristic Approaches For High-Dimensional Multi-Class Text Classification | |
| dc.type | Resource Types::text::thesis::doctoral thesis | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |